Forward-Looking Effects of Trade Liberalization

between Japan and ASEAN Members: In the

Framework of Ramsey-Cass-Koopmans Type

Economic Growth∗

Kazuhiko Oyamada†

December, 2003 (Revised May, 2004)

Abstract

This report clarifies the potential impact of trade liberalization between Japanand ASEAN members, using a forward-looking, multi-regional, multi-sectoralApplied General Equilibrium (AGE) model of global trade. We perform simu-lations with the model to evaluate several trade-liberalization programs underprospective consideration by the Japanese government, and offer some insightsinto the dynamic side of the impact. The simulation results revealed that:(i) trade liberalization between Japan and ASEAN4 has a tendency to causetrade diversion into the union followed by steady capital accumulation, whilethe case between China and ASEAN4 changes the trend of interregional invest-ment flows through announcement effects that may significantly affect patternsof trade; (ii) the announcement effects and the subsequent changes in patternsof interregional investment, caused by trade liberalization between China andASEAN4, are sufficiently large, but sensitive and may be easily affected byother policy changes such as those induce steady capital accumulation throughremoval of distortions in trade markets; (iii) Japan has a possibly importantrole in the Asian region in the linking of China and ASEAN4 through close andcomplementary relationships of Japanese and Chinese industries, particularlyin the manufacturing sector, since the initial relationship between China andASEAN4 is not very strong; and (iv) the benefit that ASEAN4 would receivefrom trade liberalization with Japan might be amplified by free trade betweenChina and ASEAN4.Keywords: applied general equilibrium; economic growth; trade liberalization;

global trade.

JEL Classification Numbers: C68, D58, F15, F41, O41.

∗Preliminary Draft. The author would like to express his gratitude to Daisuke Hiratsuka(Institute of Developing Economies), Ken Itakura (Purdue University), and Terry L. Roe(University of Minnesota) for their helpful comments and suggestions.†Research Fellow, Institute of Developing Economies, Japan External Trade Organization,

and Visiting Faculty, Department of Applied Economics, University of Minnesota.

1 Introduction

Over the past decade, the global economy has become increasingly interdepen-dent and international linkages have intensified with the launch of the WorldTrade Organization (WTO). Most of the members of the WTO are also in-volved in regional economic arrangements such as the Asia-Pacific EconomicCooperation (APEC), the European Union (EU), and the North American FreeTrade Agreement (NAFTA), or bilateral Free Trade Agreements (FTAs), in theprocess of WTO negotiations. Because it often takes a long time to agree ona single guiding formula to be used by all members in making multilateral tar-iff cuts when the coverage of trade liberalization is wide, small-scale economicarrangements such as preferential trading cooperation are regarded as beingconsistent with and supportive of the formation of large-scale frameworks. Suchsmall-scale economic arrangements that precede global trade liberalization arerapidly increasing numbers.

Along with this global economic situation, the Japanese government is ex-amining the feasibility of several economic arrangements, such as respectivelywith Chile, Korea, Malaysia, Mexico, the Philippines, and Thailand. In partic-ular, when Japan and Singapore signed an economic partnership arrangementin January 2002, Japanese Prime Minister Junichiro Koizumi emphasized theimportance of forming a comprehensive economic partnership that would in-clude an FTA, between Japan and ASEAN. In order to support the initiative,a group of professionals from Japan and ASEAN members started listing sen-sitive products and evaluating anticipated impact of liberalizing trade. Theproposal prepared by the group was presented at a top-level meeting among theASEAN+3 economies held in November, 2002. While many economists believethat increased economic integration among countries has tended to increaselong-term growth rates, and that deepening openness and economic interdepen-dence through free trade and international investment capital movement maybe major factors in generating prosperity for the global economy, it may beimportant for policy makers to numerically estimate the anticipated impact ofliberalizing trade and investment when countries launch into an economic part-nership program.

In the field of trade-related negotiations, results from Applied General Equi-librium (AGE) analyses based on certain economic data sets may be informative,because the analytical models can quantify the impact of the policy changesin a highly complicated economic system and may yield concrete evaluations.Whalley (1985) is one of the earliest research papers that quantifies the meritsof alternative actions in international trade policy. It analyzes the impact ofvarious trade liberalization initiatives, using a numerical model of global trade.More recent studies have been presented by the Global Trade Analysis Project(GTAP, Hertel ed. (1997)), which continuously carries out comprehensive anal-yses on trade-related subjects and supplies consistent data that enable us toanalyze trade-related policy changes and get information for decision making.However, since the models used in these analyses are essentially based on astatic framework, the dynamic side of the impact that is important in the fieldof open-economy macroeconomics1, for instance, how patterns of savings and

1Using a two-country model with capital accumulation and assuming pooling equilibrium,Buiter (1989) analyzes the impact of various fiscal policies on trade balance. Ono and Shibata(1992) examine the effect of home country supply-side shocks on the welfare levels of both

1

investment that lie behind growth effects translate into certain patterns of cur-rent account dynamics, have not been captured in the analyses. While addingcapital accumulation process to a static model, to be a recursively dynamic orbackward-looking model, may enable us to capture some of the positive aspectsof growth, it is quite clear that a satisfactory welfare analysis requires the intro-duction of forward-looking intertemporal preferences of households (who decideon savings), as well as forward-looking intertemporal optimization of firms (whodecide on investment)2. In this regard, this study offers some insights into thepotential impact of trade liberalization, incorporating forward-looking decisionmaking of economic agents.

When one analyzes impact of trade liberalization, three types of effects canbe considered: (i) direct and basic effects of removing trade barriers; (ii) growtheffects through capital accumulations; and (iii) effect of interregional capitalmovement through changes in capital prices. The first effects are the most basicones, which are analyzed mainly in the static framework. When a group ofcountries settle free trade, import prices of products from liberalization partnersinitially fall, because of the removal of trade barriers. The falls in import pricesexpand the union member’s demand for the products, and as a result, outputprices of the products rise. This implies the improvements of terms-of-tradeagainst outside the union, so that the union members may import more whilethey export less, and be better-off.

The second effects are the growth effects, which need the dynamic frameworkto be captured. Once a certain change occurs in patterns of global trade, theimpact last long through changes in regional investment that would be accumu-lated as capital. If capital accumulations in union members are accelerated bytrade liberalization, the members may expand productions comparative to thecase that there is no policy change takes place, forever. We may not miss thiskind of effects that continues for a long period.

The third ones are the effects that we would like to emphasize in this study,since forward-looking framework is needed. When one expects future prosper-ities of countries, those who are going to settle free trade several years later,what will she/he do? From the view point of an investor, one may invest to thecompanies in the countries involved in the liberalization program. The key fac-tor in this kind of effects is capital price, which affects patterns of interregionalcapital flows. Note that this kind of impact may arise even before the imple-mentation of the liberalization program, and affect the global trade patterns ofreal goods through current account dynamics.

The purpose of this study is to answer the following questions, using a

home and foreign countries. Some studies make efforts to extend the 2×2×2 Hecksher-Ohlinmodel into dynamic frameworks. Fisher and Vousden (1997) analyze the growth effects ofcustoms unions and free trade areas. Ono and Shibata (1991) incorporate intertemporallyoptimizing agents and analyze the effect of fiscal policy on each country’s welfare. Fromthe view point of economic growth, Islam (2001) clarifies, using a numerical multi-regionalgrowth model, that patterns and trends in the global economy are consistent partially withthe stylized facts of growth of national economies, and that optimal growth rates, optimalstructure, and growth dynamics without and with convergence constraints appear to be thesame.

2Francois et al. (1997) and Keuschnigg and Kohler (1997) address the importance ofaccumulated growth effects of trade-related policy changes. Devarajan and Go (1998) pointout the contradiction of the assumption in recursively dynamic models that the same agentbehaves rationally for one set of decisions (within-period decisions) but irrationally for another(intertemporal decisions).

2

forward-looking, multi-sectoral, multi-regional AGE model of global trade:

1. How does trade-liberalization program between Japan and ASEAN mem-bers affect regional welfare levels?

2. What are the dynamic profiles of potential impact?

We approach these questions simulating announced implementation of trade lib-eralization between Japan and ASEAN members, in comparison with the caseof China and ASEAN countries. The model used in this paper solves for a set ofinter- and intra-temporally consistent prices. Both savings and investment arethe result of dynamic optimization based on future prices that are consistentwith the achieved levels of savings and investment. Since the forward-lookingmodel is calibrated on the assumption that the benchmark data are obtainedfrom the global economy in a stationary state, we focus on the qualitative dy-namic impact of policy changes. In the evaluation of policy options, we compareimpact quantifying deviations from the values of variables given in the referencerun.

In the following section, we introduce a simple enough model to grasp againthe basic impact of changes in trade barriers that may aid us understandingthe simulation results. In section 3, we outline the major assumptions andthe structure of the forward-looking, multi-sectoral, multi-regional AGE modelused in this study. In section 4, we perform four simulations with the modeland interpret the results. Finally, section 5 concludes the paper.

2 Economic Intuition

Since the interpretation of simulation results requires reference to theoreticalanalysis, we revisit in this section some basic aspects of the effects of changes intrade barriers on the global economy, using a simple enough two-region model3.With this simplest model, we may analyze static impact of trade liberalizationon the relationship between union members and non-members, and on the rela-tionship within union members. In addition, an extension of the static analysismay shed light on recursively dynamic effects. Through this analysis, we areable to estimate the direction of impact of a certain policy changes, and alsounderstand that the major source of the effects is pre-determined by the initialcondition of economies, such as the shares in trade flows and the levels of tradebarriers.

Focusing on cross hauling, the presence of regions both importing and ex-porting the same product at the same time, we make the following assumptions:(i) domestically produced and imported goods are imperfect substitutes (theArmington assumption); (ii) domestically produced goods sold on the domes-tic market are perfect substitutes for goods sold on the export market; (iii) theeconomies determine quantities of imports according to the endogenously deter-mined output prices (perfect competition); (iv) instead of putting a balance oftrade constraint, the foreign capital inflows/outflows to balance current account;and (v) aggregate production which is analogous to factor input is fixed4.

3de Melo and Robinson (1989) analyze effects of other kind of environmental changes inthe international market using an one-small-country model.

4If we assume that the output is produced by only one factor and the production function

3

The consumer in Region r maximizes her/his composite consumption subjectto the budget constraint. The consumption aggregator function is assumed tobe convex to the origin. The consumer’s problem is:

Max Cr = Cr(TF rs)

s.t.∑s

(1 + τrs)PW sTF rs = PW rZr +∑s

τrsPW sTF rs + FS r (1)

where:

Cr is composite consumption to be maximized in Region r,

TF rs is trade flow from Region s to Region r,

Zr is exogenously given output in Region r,

PW r is price of output produced in Region r,

FS r is foreign savings in Region r that represent foreign capital inflow/outflow,

τrs is protection rate in Region r levied on trade flow from Region s, and

Cr(·) shows the consumption aggregator function.

Note that, when r = s in TF rs, trade flow includes domestically producedgoods and intra-regional trade. As a result, one can analyze effects of changesin protection rate, levied on the intra-regional trade, on both the union memberand the region outside the union with this model.

Since the global total of the foreign savings, which is given by summingup FS r with respect to r, becomes zero, we convert the consumer’s budgetconstraint (1) to the following.

∑r

∑s

(1 + τrs)PW sTF rs =∑r

(PW rZr +

∑s

τrsPW sTF rs

)(2)

In this case, the reciprocal of the Lagrange multiplier λ on Equation (2) trans-lates into the global average of protection-inclusive import prices5, (1+τsr)PW r,and Equation (2) can be dropped from the system of the model setting λ tounity as the numeraire6. Every price of output produced in Region r, PW r, isdetermined relative to the numeraire.

The first order condition for the optimum and market clearing conditionmake the entire system of the simplest two-region model. Those are:

(1 + τrs)PW s =∂Cr∂TF rs

(3)

and ∑s

TF rs = Zs (4)

is linear function of the factor input, the aggregate production can be interpreted as factorendowment. Nothing changes in results from the analysis.

5The protection-inclusive import prices are simply defined as the output prices with pro-tections.

6In the forward-looking framework, this equation defines the capital market equilibrium inevery period.

4

Assuming the consumption aggregator function is of Cobb-Douglas type, forsimplicity, the model solves six endogenous variables, TF 11, TF 12, TF 21, TF 22,PW 1, and PW 2. The solution of the model gives these endogenous variablesdescribed with share parameter αrs7 and the initial protection rate τrs.

2.1 Static Effects of Trade Liberalization on the Relation-ship between Union Members and Non-Members

Firstly, define the following three representative variables with the parametersand the initial protection rates for simplicity:

A ≡ α11α22 − α12α21 (> 0) (5)

X ≡

(α11

1+τ11

)α11A(

α121+τ12

)α12A

(α21

1+τ21

)α11A(

α221+τ22

)α12A

> 0 (6)

Y ≡

(α11

1+τ11

)α21A(

α121+τ12

)α22A

(α21

1+τ21

)α21A(

α221+τ22

)α22A

> 0 (7)

Normally, A is positive, since the shares of domestically produced goods plusintra-regional imports, such as α11 and α22, are greater than those of interre-gional imports, such as α12 and α21, in the total demand of a region. Therefore,for convenience, we assume positive A in the argument made in the rest of thissection.

Let us assume countries included in Region 1 liberalize trade with the othercountries included in the same region. This can be expressed as the reduction ofprotection rate, τ11, when we focus on the relationship between union membersand non-members.

Assuming positive A, we obtain: ∂PW 1∂τ11

< 0; and ∂PW 2∂τ11

> 0. This impliesthat, when the rate of protection levied on the intra-regional trade is reduced,output price of the product of the region where settle free trade rises, and outputprice of the region outside the union drops. As a result, terms-of-trade of theunion member improves against non-members, and the region included in theunion may import more from non-members, exporting the same amount beforethe policy change.

If we look at the impact on flow quantities of trade and consumption, it seemsmore complicated. Differentiating quantity variables in the model with respectto τ11, we obtain: ∂TF11

∂τ11T 0; ∂TF12

∂τ11T 0; ∂TF21

∂τ11S 0; ∂TF22

∂τ11S 0; ∂C1

∂τ11T 0;

and ∂C2∂τ11

S 0; when X S Y . These show that the direction of changes are thesame in the group of TF 11, TF 12, and C1, and opposite in the group of TF 21,TF 22, and C2. When the case that X is greater than Y , trade flows amongunion member and into the union from the outside increases when τ11 becomessmaller, and as a result, consumption level of the coalition member becomes

7The Cobb-Douglas type consumption aggregator is assumed to be linearly homogeneousthat the summation of αrs with respect to s is unity.

5

higher. Contrary, reduction of τ11 lessen the trade flows from the members tonon-members and intra-regional trade outside the union. Consumption outsidethe union may decrease. This implies that trade-diversion to the union membersoccurs.

2.2 Static Effects of Trade Liberalization on the Relation-ship among Union Members

Let us assume a country/region liberalize trade reducing protection againstanother country/region. This can be expressed as the reduction of protectionrate, τ12 or τ21, in the model.

As in the previous case, we obtain: ∂PW 1∂τ12

< 0; and ∂PW 2∂τ12

> 0, assumingpositive A. In the framework with the Armington assumption, cutting protec-tion raises output price of the home country/region, and pulls down the priceof the foreign output.

In turn, if we look at the impact on trade flows and consumption, we obtain:∂TF11∂τ12

T 0; ∂TF12∂τ12

T 0; ∂TF21∂τ12

S 0; ∂TF22∂τ12

S 0; ∂C1∂τ12

T 0; and ∂C2∂τ12

S 0; when

X S Y . These show that liberalizing trade has similar effects on the foreigncountry/region to on the non-members in the previous case.

When the case that both home and foreign countries/regions settle free trade,opposite impact on trade flows and consumption is to be cancelled out, anddirection of the total effect is determined depending on the initial condition ofboth economies. Output price of a country/region that has smaller share ofdomestically produced goods and intra-regional imports may have a tendencyto be more sensitive and higher than that of another country/region that haslarger share of domestically produced goods and intra-regional imports.

2.3 Recursively Dynamic Effects

Regarding the impact of the reduction of τrs on consumption as that on finaldemand, we may extend the analysis to recursively dynamic effects. Assum-ing monotonic increasing relation in final demand and investment8, investmentin Region r increases when the level of Cr becomes high. The increase of in-vestment accelerates capital accumulation in Region r, and as a result, theproduction volume of the region becomes larger in the next period. This can beexpressed as the increase of the output Zr.

If we consider the case that the volume of Z1 changes, we obtain: ∂TF11

∂Z1S 0;

∂TF12

∂Z1S 0; ∂TF21

∂Z1T 0; ∂TF22

∂Z1T 0; ∂C1

∂Z1S 0; and ∂C2

∂Z1T 0; when X S Y .

These show that when the case that X is greater than Y , effects of reductionin a particular protection rate may be amplified through the growth effectsin the subsequent periods, while the changes in the volumes of output do notaffect prices in the framework with Cobb-Douglas type consumption aggregator.Note that, if we assume Constant Elasticity of Substitution (CES) function withgreater elasticity than unity, the gains from trade may also be magnified.

8This corresponds to the cases that split savings from household’s income using constantsaving rate.

6

3 The Model

In this section, we outline the major assumptions and the structure of theforward-looking, multi-sectoral, multi-regional AGE model used in this study.We also present the basic structure of Benchmark Data and the parameteriza-tion of the model.

3.1 Major Assumptions

First, let us show the nature of the model used in this study.

Multi-Sectoral, Multi-Regional Growth Model The framework is thatof a dynamic multi-sectoral, multi-regional growth model, which is based onthe Ramsey-Cass-Koopmans type of optimal growth theory. The global econ-omy is divided into five countries/regions: (i) Japan; (ii) China; (iii) the AsianNIEs (Hong-Kong, Korea, Singapore, and Taiwan); (iv) ASEAN4 (Indonesia,Malaysia, the Philippines, and Thailand); and (v) the Rest of the World (ROW).While only four countries are included as the ASEAN members, because themodel is planned to be extended to include Foreign Direct Investment (FDI) inthe future9, the other members are included in ROW. Industries are aggregatedinto three sectors: primary industries, manufacturing, and services10. Economicgrowth is led by the exogenous growth of labor input and Total Factor Produc-tivity (TFP). In order to obtain a steady growth path as the base case, theeconomic growth rate should be equal among regions. While it is unrealistic toassume identical economic growth rates for each region, we assume the growthrates to be zero in order to focus on the qualitative dynamic impact of policychanges.

Perfect Competition The model is essentially based on the neo-classicalgrowth theory, and its solution can be regarded as the result of perfect compe-tition11. This is one straight-forward implication from the model. Since perfectcompetition is hardly realizable in actual economies, the simulation results maybe interpreted as giving only a potential picture of a hypothetical economy un-der conditions of perfect competition, on the basis of which we can abstractfundamental determinants of economic growth. When one assumes monopolis-tic or oligopolistic scale economies in the model, the impact of policy changesmay be amplified. In this regard, it can be said that results from the simulationswith the model used in this study depict sorts of lower-bound estimates.

Primary Factors The labor force is assumed to be immobile beyond the re-gional boundaries. In contrast, investment capital flows across countries/regions

9Available data that captures interregional FDI flows and stocks are limited to those fourcountries out of the ASEAN members.

10The main purpose of this sector disaggregation is to capture the spillover effects throughintermediate transactions. Sector specific growth patterns, such as those Roe and Saracoglu(2004) emphasize the importance, are not considered in this study.

11Because of the difficulties in parameterizing the model, neither imperfect competition norbiased information is incorporated in this study. While Yeldan and Roe (1994) point out theimportance of modeling non-competitive or missing market structures and heavily politicized,regulated managerial practices that are often based on imperfect and biased information, weconcentrate on an analysis assuming perfectly competitive markets.

7

(foreign capital inflow/outflow), and its flow is determined so as to balance eachcountry/region’s current accounts. It is assumed that the representative con-sumers in every country/region receives factor income from domestic firms, andthat they then invest a fraction of their income through the interregional capitalmarket. In addition to these, note that full employment of labor is assumed andplays an important role in performing simulations. Itakura et al. (2003) sug-gest that the investment capital may flood into particular regions in the wake oftrade-related policy changes with the models assuming full employment. Sinceinterregional investment capital movements affect welfare levels much in thegeneral equilibrium framework, the simulation results may present extreme pic-tures.

Exchange Rates Exchange rates for the currencies of individual countries/regionsdo not enter into the equations in the model. In a monetized extension of themodel, an explicit function of demand for money in each country/region is speci-fied and a particular regional money stock determines the monetary equilibrium.Such a specification, however, will reveal the classical dichotomy between realand monetary phenomena, as often presented in neo-classical macroeconomictheory. This dichotomy implies that behavior on the real side of the economyis independent of monetary conditions and that the monetary side alone deter-mines the price of money in terms of goods. Relative commodity prices thereforeremain unchanged if the money stock changes, and the price level is determinedby the money stock alone once real-side behavior is determined. Since the modelused in this report is a real-side trade model, the issue of the determination ofexchange rates does not arise.

Dynamic Consistency The agents’ intertemporal behavior is assumed to berational, so that the entire system of prices over time is internally consistent.This is because the model calculates variables of all the periods at the same time.Consumption and investment are determined on the basis not of what happenedin the past, but of the assumed future conditions of technology, preference, andpolicy change. Changes in the future exogenous variables can affect presentendogenous variables.

Discrete Time Formulation For the purpose of numerical implementation,the intertemporal problem is formulated in discrete time. Discounting in dis-crete time requires a dating convention. In order to keep the derivation andcalibration simple, all transactions are assumed to take place at the end of theperiod (while decisions are made or planned at the beginning of the period),following Devarajan and Go (1998).

3.2 Basic Structure

In turn, we present the basic structure of the model used in this report, focusingon the dynamic side of the model. The model is an extension of a typical staticglobal trade model, such as that presented by Hertel ed. (1997), with forward-looking properties, such as those introduced by Devarajan and Go (1998). Inthe following, subscripts r and t denote countries and time period, respectively.

8

Enterprise There is one competitive enterprise in each sector for every coun-try/region, which produces one kind of product. Production and factor inputsare all determined endogenously so that resources are optimally used from theviewpoint of a maximization of net income. Factor substitutability is assumedamong labor, capital, and intermediate input. Note that we assume that nestedfactor inputs in the production and technologies in all sectors exhibit constantreturns to scale. Given the initial capital stock, interregional rate of return andprices of primary factors, composite intermediate good, composite investmentgood, and output, the dynamic decision problem of the enterprise is to choosea time path of investment that will maximize the value of the firm, definedas the discounted sum of temporal net cash flow yielded in every period. In-vestment comprises raw capital, and is equipped to form the capital stock ofeach country/region. Since this is a long-term model, inventory is included ininvestment.

An enterprise’s optimization problem can be expressed as follows:

Max VF jr =∞∑t=1

(t∏

v=1

11 + RI v

)Rjrt

s.t. Kjrt 5 Ajrt−1 + (1− δjr)Kjrt−1 ⊥ PK jrt (8)Yjrt 5 CD(Kjrt, Ljrt) ⊥ PY jrt (9)Zjrt 5 CES(Yjrt,QH jrt) ⊥ PZ jrt (10)

limt→∞

(t∏

v=1

11 + RI v

)PK jrtKjrt = 0 (11)

where:

VF jr is the value of the j-th firm in Region r,

Zjrt is j-th gross output in Region r,

Yjrt is j-th value-added in Region r,

Kjrt is j-th capital stock in Region r,

Ljrt is j-th labor input in Region r,

QH jrt is j-th composite intermediate input in Region r,

Ajrt is raw capital installed to be j-th capital in Region r,

PK jrt is price of j-th capital in Region r,

PY jrt is price of j-th value-added in Region r,

PZ jrt is price of j-th output in Region r,

RI t is interregional rate of return,

Rjrt is current net cash flow (the subtraction of costs and investment fromsales) of j-th enterprise in Region r,

δjr is physical depreciation rate for j-th capital in Region r,

CD(·) shows the function is of Cobb-Douglas type,

9

CES(·) shows the function is of CES type, and

⊥ shows the counterpart relation between an inequality and a positive variable.

Equation (11) is the transversality condition, which places a limit on borrowingand ensures that the maximand is bounded.

Following Uzawa (1969), we assume the relation between quantity of invest-ment and installable new capital, respectively per unit of capital stock, is givenas:

AjrtKjrt

= f

(QF jrt

Kjrt

)(12)

where:

QF jrt denotes composite investment good used by j-th enterprise in Region r.

The adjustment cost function f(·) satisfies: f(0) = 0; f ′(0) = 1; f ′ > 0;and f ′′ < 0, and Equation (12) shows that adjustment cost is needed to set upinvestment goods to be installed as capital, and the cost of one unit of investmentdeclines when capital accumulation proceeds. This implies that rapid capitalaccumulation needs more capital installation cost, and as a result, desired levelsof capital stock are attained gradually with instantaneous changes in the rateof return.

First order conditions derived from this enterprise’s optimization problemformulates the investment side of the dynamics in the model. Using the condi-tions, we obtain the following relation for interior solutions:

11 + RI t+1

PK jrt+1Kjrt+1

= PK jrtKjrt −{

1− f ′(

QF jrt

Kjrt

)}PQF rtQF jrt

− (PY jrtYjrt − PLjrtLjrt − PQF rtQF jrt)

(13)

where:

PLrt is price of labor in Region r, and

PQF rt is composite price of investment good in Region r.

The second term of the right hand side is the adjustment cost for capital in-stallation, and the third term is net cash flow payable to owners of capital asdividend. Equation (13) shows that future rise of capital price affects in thedirection of increasing investment considering the adjustment cost, through re-taining operating surplus inside the enterprise.

Household Given the interregional rate of return, composite price of con-sumption good, and regional wealth, the representative consumer in each coun-try/region chooses a time path of savings that will maximize her/his discountedutility of the temporal sequence of aggregated consumption. The utility func-tion is homogenous and additively separable with constant elasticity of marginalutility. The utility is discounted by the consumer’s positive and constant rateof time preference. Since the financial claims are perfect substitutes ex ante, we

10

cannot uniquely determine the individual consumer’s optimal portfolio shares.However, since the goods are imperfect substitutes, interregional capital mar-ket equilibrium conditions define the foreign borrowings/lending for each regionendogenously. The model treats capital flows as equal to the balance of trade,adjusted for debt-service payment/receipt, and the stream of debt-service pay-ment/receipt arising from an increase in foreign borrowings/lending is incor-porated into the household’s decision making. Without uncertainty and withefficient capital markets, financial assets among countries/regions earn the sameanticipated rate of return.

A household’s optimization problem can be expressed as follows:

Max Ur =∞∑t=1

Lrt

(1

1 + ρr

)t 11− σr

(QC rt

Lrt

)1−σr

s.t.∞∑t=1

(t∏

v=1

11 + RI v

)PQC rtQC rt 5Wr ⊥ λHr (14)

where:

Ur is utility level in Region r,

Lrt is exogenous labor supply in Region r,

QC rt is composite consumption in Region r,

PQC rt is composite price of consumption good in Region r,

λHr is the marginal utility of wealth in Region r,

Wr is regional wealth (the discounted sum of current net factor income andforeign debt/asset) in Region r,

ρr is subjective discounted rate in Region r, and

σr is inverse of the elasticity of intertemporal substitution in Region r.

Similar to the enterprise’s investment, first order conditions derived from theabove optimization problem formulates the consumption side of the dynamicsin the model. When we ignore the regional growth of labor for simplicity, weobtain the following relation for interior solutions:

QC rt+1

QC rt

={(

1 + RI t+1

1 + ρr

)PQC rt

PQC rt+1

}1/σr

(15)

Equation (15) implies that higher rate of return makes future consumptioncheaper, so that future consumption increases. The rate of return RI t is de-termined by the opportunity cost of savings, which in this study is the cost offoreign borrowings.

Interregional Trade The product of the firm in every country/region is nottreated as homogeneous across countries but as imperfect substitute for thatof another. By way of example, American and Japanese cars are not treatedas a single homogeneous product (cars), but as differentiated products between

11

which there is a specific elasticity of substitution due to demand. This assump-tion is called the Armington assumption (Armington (1969)) and is necessaryto accommodate cross hauling (the phenomenon of a country both importingand exporting the same product at the same time). This is inconsistent withthe traditional Hecksher-Ohlin trade model, which is based on the premise ofhomogeneous products. The model adopts a transaction system similar to theGTAP model, presented by Hertel ed. (1997), to note the interregional trade.

Equilibrium Conditions To arrive at a solution, both the intertemporal andwithin-period general equilibrium conditions have to be satisfied simultaneously.At every point in time, the usual general equilibrium conditions require that: (i)material balance in the demand and supply of all goods in the economy holds;(ii) the demand for total labor equals its supply; (iii) government’s tax revenueis allocated between public expenditure and saving; and (iv) global-wide totalof savings equals total investment12.

The intertemporal conditions ensure that future prices and quantities arefully anticipated and factored into the behavior of investment and consump-tion. They also guarantee that the path towards a new stationary state isunique. A sufficient condition is that the discount rate and the rate of timepreference are positive and greater than the balanced-growth rate by the ter-minal period. To solve a growth model that has an infinite time horizon, wefollow the usual procedure of imposing stationary state conditions at some fu-ture terminal period. The converting procedure used in this study will be shownlater. On the investment side, the required condition is that investment is equalto the physical depreciation rate plus exogenously given post-terminal growthrate times capital stock. At the same time, current account is in equilibriumthat debt-service payment/receipt is equal to the net exports/imports13. Sincethe stream of debt-service payment/receipt is incorporated into the household’sdecision making, as noted in the part of the household’s utility maximization,this condition functions like the so-called No Ponzi-Game condition for the con-sumer’s dynamics. As long as the terminal conditions are satisfied, the sums ofvarious infinite series pertaining to the investment equation and the consump-tion function will be finite and well defined.

Conversion of the Infinite Time Horizon Model to a Computable For-mat with a Certain Terminal Period Since we know the value of the j-thfirm in Region r at every time period t is equivalent to the value of capitalPK jrtKjrt, the previously appeared VF jr can be divided into two terms:

VF jr =T∑t=1

(t∏

v=1

11 + RI v

)Rjrt

+

(T∏t=1

11 + RI t

)(1 + γr)PKT jrKjrT

(16)

12This condition can be dropped because of the Walras’ law. The counterpart variable ofthe condition (iv) is interregional rate of return. Every agent in every country/region facesthis identical rate of return in the model.

13Following Ono and Shibata (1992), exports and imports do not balance in a stationarystate in this study.

12

where:

PKT jr is price of j-th capital in Region r at the time period t = T , and

γr is post-terminal balanced growth rate in Region r.

The second term of the right hand side of Equation (16) is the present value ofthe j-th firm in Region r at the time period t = T , which must be zero whenT →∞ as the transversality condition. This value of capital is returned to thehousehold at the end of the time period t = T , and finance the series of finalconsumption after the time period t = T + 1 as the non-human wealth alongwith the human wealth.

Introducing a stationary state condition instead of the transversality condi-tion, the previously shown enterprise’s optimization problem can be convertedto a problem, which maximizes Equation (16) as the objective function subjectto Inequalities (8), (9), (10), and (17):

(γr + δjr)KjrT 5 AjrT ⊥ PKT jr (17)

Let us ignore the several kinds of income taxes, and government and foreignsavings to make discussions simple. Then, the regional wealth Wr in the house-hold’s budget constraint (14) can be expressed by both non-human and humanwealths as follows:

Wr = VF jr +∞∑t=1

(t∏

v=1

11 + RI v

)∑j

PLrtLjrt (18)

Note that the transversality condition for the household’s maximization problemis implicitly included in the above Equation (18).

Assuming that the series of consumption in the post-terminal period is fi-nanced by the value of capital, which is returned from enterprises at the endof the time period t = T , and human wealth earned in the post-terminal pe-riod, the budget constraint for the household’s final consumption (14) can beconverted to:

T∑t=1

(t∏

v=1

11 + RI v

)PQC rtQC rt

5T∑t=1

(t∏

v=1

11 + RI v

)∑j

(Rjrt + PLrtLjrt)

(19)

Other optimal conditions derived from the previously shown household’s utilitymaximiztion problem are not affected by this conversion.

Choice of the Terminal Period As variables of different time-periods are in-terdependent, the computation burden is much larger than that for models thatcalculate solutions period by period (recursively dynamic or backward-lookingmodels). Moreover, extensions of the calculation horizon increase calculationdifficulty more than proportionally, and expansions of models with respect tothe number of sectors or regions are more difficult. Because of these difficultiesand a limited amount of computational resources, we set the terminal period atT = 5014.

14The qualitative changes are not affected by the choice of terminal period. See Devarajanand Go (1998).

13

Software The model is formulated as a Mixed Complementarity Problem(MCP) and solved by “PATHC” of the General Algebraic Modeling System(GAMS)15. MCP is a set of Kuhn-Tucker conditions derived from certain opti-mization problems.

3.3 Overview of the Benchmark Data

The source of data for the model is the GTAP version 5 database. A fixed pro-portion of services output is supplied for interregional shipping services. Coun-tries/regions and industries are aggregated into five and three, as noted in theprevious section. Assuming that the data is obtained from an economy in astationary state, parameters and exogenous variables are calibrated from thedata. In this section, we document the GTAP version 5 database that formsthe basis for our analysis.

Basic System of the GTAP Data The GTAP database is a set of regionalinput-output tables and sectoral trade flows that connect sectoral exports andimports that appear in the input-output tables, plus several kinds of estimatedelasticity. The target year is 1997. There are four sheets of trade-flow data,which are respectively presented at wholesale prices, F.O.B. prices, C.I.F. prices,and protection-inclusive market prices. The differences among these four sheetsconsist of ad valorem equivalent domestic transportation margins and exportsubsidies, international shipping margins, and import tariffs, import quotas,antidumping duties and non-tariff barriers. Note that we collectively handlethe latter four (import tariffs, import quotas, antidumping duties, and non-tariff barriers), as a single item to be removed in the simulations performed inthis report.

Service Trade Service trade includes trade in factor services (interest, profits,and dividends) as well as trade in non-factor services (business, insurance, andfinancial services). In the GTAP framework, we have data only on non-factorservices trade, which are in turn broken down into shipping and non-shippingservice components. In the model, a fixed fraction of output is supplied forinterregional shipping services. Since the GTAP database does not fully includeestimated trade barriers, which might exist in service trade, our simulationsdo not reflect reforms in the service sector. Francois (2001) estimates ad val-orem equivalent distortions in trade in services, and incorporates mark-ups bymonopolistic firms. One point is that this kind of modifications may vastly mag-nify the anticipated impact of trade liberalization. While the real service sectormay remain restrictive, it is not included in this study because of difficulties inmeasuring the distortions or mark-ups: therefore, the simulation results in thisstudy may be regarded as lower-bound estimates.

Behavioral Parameters Some of the behavioral parameters used in thisanalysis, such as the set of substitution elasticity for the CES aggregators, areweighted average of the values provided by GTAP database. The substitutionelasticity for the commodities from different countries/regions are on the as-sumption of so-called Rule of Two. Other parameters are calibrated from the

15Brook et al. (1992).

14

benchmark data to reproduce the initial equilibrium in a stationary state16, andthen make the global economy to achieve balanced steady growth. This steadygrowth path is regarded as the reference run, which is used to scale the impactof policies simulated. As shown in Section 2, the behavioral parameters rule thesimulation results, while the formulation of the model also dominates the resultsin AGE analyses. No individual component can be considered more importantthan the others. For some simulations, it is the accounting identities that de-termine results, whereas the behavioral parameters may play a relatively smallrole. For other simulations, the specification of certain elasticity is of paramountimportance.

4 Simulations

We now report on the results of four simulations, categorized into two types,performed in this forward-looking framework. The two categories distinguishsimulation scenarios according to whether the primary industries are includedor not included in the trade-liberalization program. This is because some groupsare feeling concern for the case that liberalizing trade in primary industries,especially in agriculture, reduce domestic production volumes of the sector whenthey face a more competitive trade market.

In the case in which trade liberalization is implemented for all of the sectors(let us call this as “Case A” in the simulations), we consider three scenarios17:(i) trade liberalization between Japan and ASEAN4 that takes place in the fifthperiod; (ii) trade liberalization between China and ASEAN4 that is imposed inthe fifth period; (iii) sequential implementation of (i) and (ii) whereby China(initiating free trade in the fifth period) precedes Japan (liberalizing trade inthe ninth period). The third scenario considers the present situation that Chinais going ahead of Japan in preparing to conclude agreements on economic part-nerships with ASEAN members18.

In turn, we consider one additional scenario, which corresponds to (i), tradeliberalization between Japan and ASEAN4 that takes place in the fifth period inCase A, and in which trade liberalization is implemented for the manufacturingand service sectors (Case B). The reason is that there is only a small differencebetween Cases A and B in other combinations of policies.

In these experiments, we examine announced effects of the policy changesthat are fully anticipated four or eight periods ahead, and trade barriers againstthe other member countries/regions of the union are removed forever after lib-

16A detailed calibration procedure in a forward-looking model is clearly presented by De-varajan and Go (1998).

17In the simulations, Singapore, who is grouped in the Asian NIEs in this analysis, is notincluded in trade-liberalization programs. This is because, as mentioned before, the modelused in this study is designed to be comparable with an extended model that incorporatesFDI, which is now under construction. Since we make a great point of analyzing qualitativedynamic impact of liberalizing trade in this study, we prioritized comparing results fromseveral forward-looking models, which have identical aggregation level but are different in thechoice of assumptions over simulating plausible policy changes.

18There is no special reason in our choice of the period when policy changes occur. Tohighlight the impact of two trade-liberalization programs, assuming sequential implementationmakes things clear rather than assuming simultaneous one. It is difficult to split effects of apolicy change from those of another in simulating several policy implementations in the sameperiod.

15

eralizing trade. Note that, in the figures, the deviations in values of variablesfrom the base case are shown.

4.1 Existing Trade Barriers

First of all, we identify the trade barriers to be removed in the simulations. Table1 shows the ad valorem equivalent protection rates for 1997 levied on sectoraltrade flows from the source countries/regions (appearing in the top row) to thedestination (appearing in the left column). The values of protection are obtainedby subtracting the trade flows at C.I.F. prices from those at protection-inclusivemarket prices. Specifically, these margins include import tariffs, import quotas,antidumping duties, and non-tariff barriers. Note that we collectively handlethem in the simulations as the trade barriers to be removed.

Japan China NIEs ASEAN4 ROWJapan Primary 0.000 13.900 11.360 2.191 12.742

Manuf. -1.086 8.477 4.933 8.974 9.239Services 0.000 -0.001 -0.007 -0.002 -0.004

China Primary 9.492 0.003 9.757 4.778 17.767Manuf. 15.632 0.001 16.489 19.150 14.111Services -0.019 0.000 -0.010 -0.016 -0.014

NIEs Primary 10.121 49.825 9.744 3.753 11.586Manuf. 4.224 3.568 1.968 1.685 4.188Services 0.027 0.018 0.101 0.136 0.083

ASEAN4 Primary 12.038 14.794 36.071 6.613 6.043Manuf. 11.116 11.771 8.388 9.998 7.812Services 0.000 0.000 0.000 0.000 0.000

ROW Primary 8.379 9.751 12.881 9.111 4.588Manuf. 5.651 8.907 6.744 7.687 4.586Services 0.147 0.131 0.395 0.164 0.166

Table 1: Ad Valorem Equivalent Protection Rates (%, 1997)

4.2 Dynamic Effects

In this subsection, we focus on the dynamic impact of the four scenarios of trade-liberalization programs in the Asian region. In a static framework, output pricesof the commodities produced in countries/regions involved in a union tend torise relative to the global average of output prices, as seen in Section 2. Thehigh prices of products of union members in comparison with those of non-members improve the terms-of-trade, and enable the members to be better off.Under the condition of low price distortion and resulting more efficient intra-and interregional resource allocation, trade diversion may occur, and as a result,economic volumes of union members may enlarge in the global economy throughexpansions of production among the members. Such growth effect amplifies thestatic impact through capital accumulations in a dynamic framework.

In the framework of forward-looking dynamics, capital price becomes onekey factor. Changes in capital prices triggered by the future static shock will

16

lead movements into patterns of interregional investment flow in the period be-fore the policy change, following the movements in interregional trade in goods.The effects of the movements in interregional investment are crucial for theglobal economic situation. When a policy change, such as trade liberalization,is announced to be implemented in a certain future period, the expected staticimpact and subsequent growth effects raise the capital prices19 among the unionmembers relative to those among the non-members. Thus, the effects of futuretrade liberalization appear in the periods before policy changes through themovements in interregional investment flows, and the existence of price distor-tions characterizes the impact before and after policy implementation. In thepre-implementation period, changes in capital prices affect the allocation of in-terregional investment, and at the same time, affect the real markets throughchanges of output prices.

4.2.1 Case A (i): Trade Liberalization between Japan and ASEAN4

We start with simulating trade liberalization between Japan and ASEAN4 thattakes place four periods ahead. In this case, the basic impact confirms oureconomic intuition obtained in the previous section. Since the values of A, X,and Y , given in Equations (5) to (7), are respectively 0.92, 63.14, and 0.05, whenwe consider the relationship between union members and non-members, outputprices of Japan and ASEAN4 rise relative to the global average of output prices.According to these improvements of terms-of-trade, both trade flow betweenJapan and ASEAN4, as well as inflow to the free-trade area from China, theAsian NIEs, and ROW increase. In contrast, trade outside the union, and tradeoutflows from Japan and ASEAN4 to other countries/regions are respectivelyreduced. Trade diversion to Japan and ASEAN4 occurs.

Figure 1 shows the dynamic impact of trade liberalization between Japanand ASEAN4 on regional averages of output prices relative to the global av-erage. The values are deviations from the base case that there is no policychange takes place, and the values above/below the zero level can be regarded asimprovements/aggravations of terms-of-trade, so that the corresponding coun-tries/regions may import more/less while exporting less/more. The previouslymentioned basic impact of liberalizing trade are appearing just after the fifth pe-riod, when Japan and ASEAN4 settle free trade, that output prices of Japaneseand ASEAN4 products rise. The basic improvements/aggravations of terms-of-trade affect capital prices to rise/fall in the period before the policy implemen-tation, and yields announcement effects that change patterns of interregionalinvestment flow. If the country/region who is involved in a trade liberalizationprogram accumulates sufficient capital, capital price in the country/region mayfall in the post-implementation period. If capital stock is not sufficiently ac-cumulated within a certain period, because of the existence of adjustment costfor capital installment, capital price may remain high. In the former case, cap-ital formation in the country/region is financed mainly by domestic investmentthrough growth effects, and the share of interregional investment forwarded tothe country/region decreases.

The following points can be observed if we look on Figures 2 and 3, whichshow export, import, and output values of Japan and ASEAN4, respectively,

19Capital prices may be translated into stock prices. Capital price times interregional rateof return in the model forms dividend and earnings retained in an enterprise to be invested.

17

18

in addition to Figure 1. First, Japanese and ASEAN4 export volumes are re-duced before liberalizing trade. This is because, the capital prices in Japan andASEAN4 rise in comparison with those of the other countries/regions before thepolicy change, so the proportion of the interregional investment flow directedtoward the region, particularly toward Japan, becomes larger. Consequently,Japan and ASEAN4 increase foreign savings by cutting exports and importingmore, in the pre-implementation period. This increased foreign savings financesthe expansion of investment in Japan and ASEAN4.

Second, there is a large difference in the time profile of output prices, be-tween Japan and ASEAN4. The average output price of ASEAN4 products fallssharply after the policy change, and the price goes below the level in the basecase after the 15th period. Consequently, impact on exports from ASEAN 4exceeds that on imports after the 15th period.

One interesting point on the fall of output price in ASEAN4 is that it comesfrom the manufacturing sector. Figure 4 shows effects on the sectoral out-put prices in ASEAN4. Output prices of manufactured products drop afterthe implementation of the policy. This indicates that liberalizing trade withJapan expands the volume of manufacturing sector in ASEAN4, concentratingresources from primary and service sectors, and resulting excess supply of theproducts lowers the output prices. While some important elements, such as in-terregional fragmentation and intra-group trade in Multi-National Enterprises(MNEs), are not modeled in this study, the close relationships among man-ufacturers operating in Japanese and ASEAN4 might be suggested with thisresult.

Another aspect for the difference in the time profile of output prices comesfrom the capital-investment side. Figures 5 and 6 show the effects on values ofconsumption, investment, and capital stock in Japan and ASEAN4. It is clearthat the investment in Japan increases before liberalizing trade, and in the post-implementation period, it rapidly decreases to the new stationary state level. Incontrast, the investment in ASEAN4 increases with the implementation of tradeliberalization then gradually decreases to the new stationary state level higherthan the base case. The deviation of investment in ASEAN4 is more than fivetimes higher than that in Japan at the highest level. This implies that a largeportion of interregional investment flows into ASEAN4 after the policy change isoutflow from Japan. The rush of interregional investment to ASEAN4 enlargesthe stock of foreign capital of ASEAN4, and a portion of the borrowings isrepaid to the creditors within the simulation period. The terminal condition forthe model used might affect this result. Since the capital in ASEAN4 is steadilyaccumulated, the foreign borrowings may easily be repaid, and the capital pricesin ASEAN4 fall along with advances of capital accumulation. This is reflected tothe gradual fall in output prices in ASEAN4 in the post-implementation period.

Figures 5 and 6 also show that the levels of investment in Japan and ASEAN4toward the new stationary state are respectively one and three percent higherthan the base case. It implies that investment in both countries/regions iscontinuously expanded by the trade liberalization. This is the result of growtheffects, in addition to the improvement in terms-of-trade. If we look at GDP, theimpact of trade liberalization between Japan and ASEAN4 is clearly illustrated.The effects on GDP are shown in Figure 7. Since the values in the figureshows the deviations from the base case, GDP levels in Japan and ASEAN4 aresuccessively raised by one and more than three percent, respectively.

19

20

While GDP may be a major indicator that captures the economic condition,increases in income do not directly imply improvement in welfare levels. Evenwhen income levels increase, welfare levels may worsen if commodity prices risemore than increases in income. In this regard, we also check the welfare levels,which are measured by the consumption quantities in each country/region.

Figure 8 shows the effects on consumption quantities of every country/regioncomparative to the case that no policy change takes place. Since the welfarelevels reflect the story noted above, Japan and ASEAN4 are better-off in thenew stationary state. Improvement in Japanese welfare is the result of theimprovement in terms-of-trade, expansion of imports from ROW with cheapprices. On the other hand, ASEAN4 enjoys consumption of imports from ROWafter the policy implementation, and also domestic products that finally be-come cheaper than the base case. After the 15th period, consumption level ofASEAN4 reaches 1 percent higher than that in the base case, and it contin-ues forever. The trough in the fifth period is caused by the sharp rise in theprices of home products supplied for domestic consumption. Since the largesttrade partner of Asian countries/regions is ROW, ROW recovers welfare levelwithdrawing interregional investment capital.

The levels of consumption also show the levels of saving. As seen, investmentby ASEAN4 sharply increases in the fifth period, and there is a change in thepatterns of interregional investment flows to be forwarded to ASEAN4. Theconsumption level of ASEAN4 shows that the saving rate in ASEAN4 is alsoraised to finance the expansion of investment within the region in the fifthperiod.

4.2.2 Case A (ii): Trade Liberalization between China and ASEAN4

The second simulation is trade liberalization between China and ASEAN4. Asin the case between Japan and ASEAN4, trade liberalization is announced fourperiods before the implementation. In this case, the story seems to be totallydifferent from the previous case. While the values of A, X, and Y are respec-tively 0.86, 83.09, and 0.08, when we consider the relationship between unionmembers and non-members, price effects are dominant in the overall impact.The rise in output prices in China and ASEAN4 spills over to the global econ-omy through intermediate use of imports. Consequently, the global average ofoutput prices is three times higher than in the base case, relative to the inter-regional rate of return. This is because the initial relationship between Chinaand ASEAN4 through trade is the weakest in the relationships among the fivecountries/regions modeled in this study. Table 2 shows the initial relationshipsamong countries/regions with regard to both exports and imports. In the table,each country/region labeled at the top is the source, and in its column for thedestinations listed at left.

It can easily be seen that the shares of China show the smallest valuesin both exports (4.26%) and imports (4.72%) of ASEAN4, and that those ofASEAN4 are similar in Chinese exports and imports (3.53% in exports and3.92% in imports). Because of initial weaknesses in trade relationships, productswith inflated prices are exported from both China and ASEAN4, and usedas intermediate inputs in countries/regions all over the world. This causes asecondary price rise in commodities produced in Japan, Asian NIEs, and ROW,and amplifies the price effects several times larger than the initial impact. As

21

Panel A: Export SideJapan China NIEs ASEAN4 ROW

Japan 0.00 17.04 8.38 13.91 5.93China 8.66 0.00 12.52 3.53 1.85NIEs 18.35 14.42 10.43 19.82 5.41ASEAN4 10.12 4.26 10.95 5.90 2.76ROW 62.87 64.28 57.71 56.85 84.05Total 100.00 100.00 100.00 100.00 100.00

Panel B: Import SideJapan China NIEs ASEAN4 ROW Total

Japan 0.00 10.15 9.82 9.23 70.80 100.00China 21.25 0.00 29.57 4.72 44.45 100.00NIEs 18.48 7.10 10.12 10.88 53.43 100.00ASEAN4 19.07 3.92 19.87 6.06 51.08 100.00ROW 6.25 3.13 5.53 3.08 82.01 100.00

Table 2: Share of Trade Partner (%, 1997)

mentioned before, one cause of this phenomenon is because of the assumptionof full employment. If one were to include the factor of redundant workers inAsian countries/regions, the price effects would be moderate.

Taking a look on regional averages of output prices shown in Figure 9, whichare relative to the global average, to see regional terms-of-trade, an interestingpoint can be observed. One may find that the average output price of Japaneseproducts takes the highest values in the global market throughout the simulationperiod. This is caused because the Japanese intermediate use of imported ma-terials from China and ASEAN4 occupies three times larger shares than that inROW. According to the strong structural relationships among Japanese indus-tries and producers in China and ASEAN4, Japan also improves terms-of-tradeand becomes prosperous along with China and ASEAN4 liberalize trade.

From the view point of interregional capital investment, the improvementof Japanese terms-of-trade raises capital price in Japan, as seen in the pre-vious simulation. Since global investment mainly flows into both China andASEAN4 during the pre-implementation period in this case, Japanese capitalprice remains high and is reflected to output prices in Japan.

An important point here is that the announcement effects are much sensitiveto some other policy changes that may affect people’s future expectation, suchas increases in consumption tax or pension fund burden that are under consid-eration in Japan, or possible free trade programs between the United States andAsian economies. It should be reminded that our experiments simulate somepure situations that do not include any other environmental change, which maytake place all around the world. Inclusion of such policy changes in addition tothe trade liberalization considered here may show different results.

The effects of trade liberalization between China and ASEAN4 that talesplace in the fifth period on export, import, and output values of China andASEAN4 are respectively shown in Figures 10 and 11. It is clear that exports ofChina do not exceed the volume in the base case while imports expand through-

22

23

out the simulation period, different from the Japanese response in the previoussimulation. In contrast, the response of ASEAN4 to trade liberalization withChina is similar to that in the case with Japan, that exports decrease beforethe policy change because of the increase of foreign capital inflow during theperiod, both imports and exports increase after settling free trade and effectson import values exceed those on export for a while because of the improvedterms-of-trade. Then, trade balance becomes positive along with the drop ofoutput prices, and ASEAN4 repays foreign borrowings.

These results show that Japanese response crucially affects trade patterns inthe Asian region, as well as there are two types of industrial relationships amongJapan, China, and ASEAN4. Chinese and ASEAN4 products are respectivelycomplementary to Japanese industries, while the relationship between Chinaand ASEAN4 is competitive. Since it is not strange that investors expect Japanto be prosperous because its productions are complementary to the productsfrom countries/regions that are going to settle free trade, capital prices in Japanrise even in the case of trade liberalization between China and ASEAN4, andeven in the period just the policy change is announced. On the other hand,the competitive relationship curbs Chinese exports and keeps output price ofChinese products high, while the output price of manufactured products madein ASEAN4 falls as in the previous case. Figure 12 shows the sectoral outputprices in ASEAN4. It is clear that the prices of manufactured goods are lessthat those in the base case.

Then, let us go ahead to the macroeconomic impact. Figures 13 and 14 showthe effects of trade liberalization between China and ASEAN4 on GDP andwelfare levels, respectively. Since Japan, China, and ASEAN4 respectively havestrong trade relationships with ROW, these countries/regions increase importsof cheap products from ROW and become better-off. The GDP volumes ofJapan, China, and ASEAN4 become more than 1 percent larger than those inthe base case. A point is that, if we look at welfare levels, improvement ofASEAN4 remains less than 1 percent, while Japan and China respectively have2 and 5 percent higher levels than in the base case. These are the result ofterms-of-trade effects, and show how large the influence of Japanese response tothe policy change, on the global economy.

One more difference from Case A (i) is that welfare levels do not surgethrough the simulation period. This implies that the growth effects are not sostrong. In this case, the initial impact of liberalizing trade between China andASEAN4 itself might not be so large; however, Japan responds sensitively tothe price increases in products made in China and ASEAN4, and this affectstrade patterns in the global economy. This means that Japan may have animportant role to play in linking with both China and ASEAN4, since theinitial relationships with them are not so strong.

4.2.3 Case A (iii): Sequenced Implementation in Which China Pre-cedes Japan in Liberalizing Trade with ASEAN4

Let us proceed to the third simulation, in which China gets a head start onJapan in liberalizing trade with ASEAN4. Both policy changes are assumed tooccur in the fifth and ninth periods, respectively, and those are fully anticipatedby all of the economic agents in the global economy. Figure 15 shows thatthe effects of sequenced implementation of two types of trade liberalization on

24

25

regional averages of output prices.Two critical differences from the previous two scenarios can be observed. One

is that average output price of goods made in ASEAN4 takes the highest valuesin the periods before the policy changes and reaches more than 1 percent higherthan the base case when Japan and ASEAN4 settle free trade. In addition,the price in ASEAN4 remains higher than the base case in the new stationarystate. The second difference is that Japanese output prices takes lower valuesin the pre-implementation period in comparison with the previous two cases. Itis symmetrical response to ASEAN4.

One possible reason to these results is that ASEAN4, who settles free tradewith two countries, hails foreign capital making investors to expect that theregion supposed to enjoy positive impact of the policy changes. The amountcorresponding to the increased foreign capital flow to ASEAN4 lessens foreigninvestment toward Japan in comparison with Case A (i).

Then, let us examine the effects on export, import, and output values ofJapan, China, and ASEAN4, depicted in Figures 16 to 18. One may find thatpattern of Chinese exports and productions differ from those which observed inCase A (ii). In this sequenced case, exports from China increase by more thanthree percent in the new stationary state level. Since Japan settle free tradewith ASEAN4 in the ninth period, output prices of Chinese products remainlow relative to the global average, making Chinese export volumes larger in theglobal market. This is an effect of Japanese trade liberalization with ASEAN4,which lower the Chinese output prices, and the positive effects help China toexpand production, especially in the manufacturing sector, under the conditionof Chinese liberalization with ASEAN4.

In turn, the impact on Japan is close to that seen in Case A (i). This isbecause the impact of liberalizing trade between Japan and ASEAN4 is muchgreater than those between China and ASEAN4, so that the effects observedin the previous simulation, Case A (ii), are concealed. It is not strange sincepeople’s future expectations, which affect patterns of interregional investmentflow and regional capital prices, change sensitively to the policy change that hasthe largest impact. In this reason, announcement effects in this case becomesimilar to those in Case A (i).

Figure 18, which shows the impact on export, import, and output values ofASEAN4 products, also supports this hypothesis. While the impact on ASEAN4seems to be simple combination of Cases A (i) and (ii), the shocks when freetrade with Japan starts are several times larger than the impact when tradeliberalization with China takes place.

Finally, let us look at the impact on GDP and welfare levels. Both Figures19 and 20 show that the benefit ASEAN4 would receive from trade liberalizationbetween Japan and ASEAN4 is amplified by the precedent free trade betweenChina and ASEAN4. This is because less distortion makes results in benefit forthe countries/regions involved in plural coalitions.

4.2.4 Case B (i): Trade Liberalization between Japan and ASEAN4Excluding Primary Industries

The last simulation enables us to verify impact of excluding primary sectorsfrom the liberalization program for Japan and ASEAN4. In comparison withCase A (i), one may find that the effects on the average output price in ROW

26

27

becomes positive in the new stationary state, which is captured in Figure 21.On the other hand, price changes in Japan and ASEAN4 are rather moderatefrom the case of full liberalization. These come from the smaller inflow offoreign capital from ROW to Japan and ASEAN4, and consequent repaymentof foreign borrowings by ASEAN4. It is shown in Figure 22, which depicts theimpact on Japanese export, import, and output values, in which the exportvolume becomes larger and import is smaller.

If primary industries are not included in the liberalization program, Japaneseimports of products of primary industries are smaller, and Japanese domesticproduction volume of this sector becomes slightly larger than in the case thatincludes this sector in the liberalization program. This implies that resourceallocation for the sectors other than primary becomes smaller in Japan.

An interesting point is that the volume of Chinese manufacturing becomessmaller and the other sectors larger, while Japanese production of the primarysector becomes larger. These are shown in Figures 23 to 26, which depict thesectoral production quantities in Japan and China in both Cases A (i) and B(i). It is well explained if we assume that the manufacturing sectors in bothChina and ASEAN4 individually have close and complementary relationshipswith Japanese manufacturers, as mentioned in the previous simulation.

When resources allocated to the manufacturing sector are reduced in Japan,intermediate use of imported materials from ASEAN4 increase in order to main-tain the sector’s production volume, because these two countries/regions settlefree trade, replacing Chinese made materials. Since China and ASEAN4 donot have so close relationship, their products may be rivalry rather than com-plementary. Therefore, Chinese manufacturing reduces the volume, and moreresources are allocated to the primary and service sectors in the country.

Other variables, such as production volumes and output prices in the othercountries/regions, are not sufficiently affected.

It is important to note that these consequences are obtained on the assump-tion that sectoral differences are characterized only by the values of parametersin production functions calibrated with the benchmark data. We did not madeany other treatment, at this time, which characterizes differences among threesectors modeled in this analysis, such as scale economics, sector specific growthpatterns, and market structures for the primary factors.

The GDP and welfare levels, which are captured in Figures 27 and 28, showoverall impact of liberalizing trade between Japan and ASEAN4 except theprimary products. As mentioned, the negative impact on the Asian NIEs andROW are reduced from the case of full liberalization, so the deviation of GDPof ROW from the base case becomes positive in the new stationary state. Incontrast, Japanese consumption decrease from Case A (i), because of the lossfrom distortions left in the primary sector, and it affects China to also reduceconsumption. It is interesting that ASEAN4 is not affected so much.

4.3 Welfare Effects

In this subsection, we measure the effects of four cases of trade liberalizationon the regional welfare, based on the idea of Hicksian Equivalent Variations(EV). EV are the amount of money equivalent to the changes that have alreadytaken place in the base case. In other words, the income changes that movethe agent to the post-change welfare levels. In our dynamic framework, we

28

29

30

calculate the discounted sum of temporal EV obtained in each period to findthe accumulated growth effects that policy changes potentially have20. Thewelfare gains from trade liberalizations are reported in Table 3. Note that thesewelfare gains are closely related to the welfare levels shown in Figures 8, 14, 20,and 28 in the previous subsection, which are the discounted integral values of theimages valued with the income in the base case. Since our EV are accumulatedthrough 50 periods to capture the results from growth effects, it is importantnot to directly compare with the EV calculated in the other analyses based onstatic AGE models.

Panel A: Millions U.S. Dollars at 1997 Constant PricesCase A (i) Case A (ii) Case A (iii) Case B (i)

Japan 168800.80 1658510.00 193797.70 94139.28China -15560.10 526476.10 46247.82 -36715.00NIEs -27887.00 -251987.00 -45111.20 -17200.50ASEAN4 73534.28 41863.28 117621.80 71160.14ROW -67290.00 -1929980.00 -151189.00 18229.82Total 131597.98 44882.38 161367.12 129613.74

Panel A: Percentage ChangesCase A (i) Case A (ii) Case A (iii) Case B (i)

Japan 0.013 0.122 0.014 0.007China -0.021 0.682 0.061 -0.049NIEs -0.020 -0.180 -0.032 -0.012ASEAN4 0.134 0.076 0.214 0.130ROW -0.002 -0.059 -0.005 0.001Total 0.104 0.641 0.252 0.076

Table 3: Welfare Gains

There are several points to be noted. First, the results from simulationswith Case A (i) and (iii) confirm the orthodox proposition that all of the coun-tries/regions that enter into free trade are better-off, while countries/regionsoutside the union tend to be worse-off. However, as we saw previously, an-nouncement effects and resulting terms-of-trade shock push the Japanese wel-fare level in Case A (ii) to ten times larger than in Case A (i). Rather thanJapanese welfare gains, a welfare loss of ROW reaches thirty times larger. Inour framework, the interregional investment flow greatly affects the global tradepatterns. One solution may be parting from the assumption of full employmentto avoid interregional capital movement to be large and remains in the same po-sition for a long period. Such movement is essentially sensitive, as mentioned,and easily affected by the other policy changes.

Second, while it also supports the results from straight-forward analyses, lessdistortion results in more benefit for the countries/regions involved in a union.If we look at Case A (iii), Japan and ASEAN4 receive the larger benefit thanin Case A (i). Instead, the welfare losses of the countries/regions outside theunion are also increased. From the view point of ASEAN4, while the welfare

20Since we calculate the discounted sum of entire changes caused by the policy changes, thevalues of regional EV become several times larger than the results obtained by static analyses.

31

gains from liberalizing trade with China are not large, the gains from free tradewith Japan might be beneficial owing to growth effects.

Third, the welfare gains of Japan in Case B (i), when primary industries areexcluded from the liberalization program, come to less than sixty percent of thegains in Case A (i), the case of full liberalization. In contrast, ROW recoversto get an overall gain in welfare in Case B (i). The increase in consumption inROW in the latter half of the simulation period is larger.

5 Concluding Remarks

The purpose of this study was to clarify the potential impact of trade liberal-ization between Japan and ASEAN members, using a forward-looking, multi-regional, multi-sectoral AGE model of global trade. The model can be used toanalyze questions where the response of intertemporal variables such as savingsand investment are important, and the structure of the global economy is alsorelevant.

Simulations with the model revealed the response of the global economy tofour types of trade-liberalization program. The key findings can be summarizedas follows:

1. Trade liberalization between Japan and ASEAN4 has a tendency to causetrade diversion into the union followed by steady capital accumulation,while the case between China and ASEAN4 changes the trend of inter-regional investment flows through announcement effects that may signifi-cantly affect patterns of trade.

2. The announcement effects and the subsequent changes in patterns of in-terregional investment, caused by trade liberalization between China andASEAN4, are sufficiently large, but sensitive and may be easily affectedby other policy changes such as those induce steady capital accumulationthrough removal of distortions in trade markets.

3. Japan has a possibly important role in the Asian region in the linkingof China and ASEAN4 through close and complementary relationships ofJapanese and Chinese industries, particularly in the manufacturing sec-tor, since the initial relationship between China and ASEAN4 is not verystrong.

4. The benefit that ASEAN4 would receive from trade liberalization withJapan might be amplified by free trade between China and ASEAN4.

There are several potentially important issues that are not taken into accountin the present analytical framework. First, the assumption that the globaleconomy is on a balanced growth path at the initial point (used in order tocalibrate the model) is unrealistic. Since the Asian economies are still in theprocess of development, it is appropriate to think that the global economy ison a dynamic adjustment path. While Lau et al. (2002) offer a procedure forallowing different regional growth rates in a stationary state, the econometricapproach still has importance in projections.

Second, since trade liberalization may affect fiscal budgets by reducing rev-enues from import and export duties, it is important to shed light on the possible

32

negative impact that trade liberalization may have. Economic growth may bedecelerated through the accumulation of public capital. The next issue of thisstudy would include such activities of the public sector.

Third, impact on the trend of FDI cannot be captured clearly in this analysis.An effort to include decision making on investment by MNEs has been made,but several important profiles of FDI still remain that are difficult to model.

Fourth, the present analytical model seems to be too sensitive to changes inpatterns of interregional investment. As Itakura et al. (2003) suggest, modelingwithout an assumption of full employment may be an important subject.

Fifth, it may also be important to include sector specific growth patterns.Roe and Saracoglu (2004) stress the importance to model higher TFP growthrate observed in agricultural sector compared to the others, along with sectorspecific factors, such as land.

Finally, it would be crucial from a political-economic standpoint to incorpo-rate economies of scale. We therefore feel that it is also important to abandonthe assumption of perfectly competitive markets in the future.

References

Armington, P. S. (1969), “A Theory of Demand for Products Distinguished byPlace of Production”, International Monetary Fund Staff Papers.

Brooke, A., D. Kendrick, and A. Meeraus (1992), GAMS: A User’s Guide.Release 2.25., Scientific Press, San Francisco.

Buiter, W. H. (1989), Budgetary Policy, International and Intertemporal Tradein the Global Economy, North Holland, Amsterdam.

de Melo, J., and S. Robinson (1989), “Product Differentiation and the Treatmentof Foreign Trade in COmputable General Equilibrium Models of SmallOpen Economies”, Journal of International Economics, 27, 47–67.

Devarajan, S., and D. S. Go (1998), “The Simplest Dynamic General Equi-librium Model of an Open Economy”, Journal of Policy Modeling, 20(6),677–714.

Fisher, E. O., and N. Vousden (1997), “Dynamic Trade Creation”, mimeo.

Francois, J. F. (2001), THE NEXT WTO ROUND: North-South stakes in newmarket access negotiations, CIES Aderaide and the Tinbergen Institute,Adelaide.

Francois, J. F., B. J. McDonald, and H. Nordstroem (1997), “Capital Accu-mulation in Applied Trade Models”, in Applied Methods for Trade PolicyAnalysis, ed. by J. F. Francois, and K. A. Reinert, chap. 12, pp. 364–382,Cambridge University Press, Cambridge.

Hertel, T. W. (ed.) (1997), Global Trade Analysis, Cambridge University Press,Cambridge.

Islam, S. M. N. (2001), “Optimal Multiregional Global Economic Growth: For-mulation of an Optimal Growth Program, Growth Prospects, and Conver-gence”, Journal of Policy Modeling, 23, 753–774.

33

Itakura, K., T. W. Hertel, and J. J. Reimer (2003), “Labor Market Impactsof the Japan-ASEAN Free Trade Agreement”, in International Conferenceon Emergent Trilateralism in the Pacific Basin, University of California,Berkeley.

Keuschnigg, C., and W. Kohler (1997), “Dynamics of Trade Liberalization”, inApplied Methods for Trade Policy Analysis, ed. by J. F. Francois, and K. A.Reinert, chap. 13, pp. 383–434, Cambridge University Press, Cambridge.

Lau, M. I., A. Pahlke, and T. F. Rutherford (2002), “Approximating Infinite-Horizon Models in a Complementarity Format: A Primer in Dynamic Gen-eral Equilibrium Analysis”, Journal of Economic Dynamics and Control,26, 577–609.

Ono, Y., and A. Shibata (1991), “Trade, Capital Accumulation, and Welfare”,ISER Discussion Paper.

(1992), “Spill-Over Effects of Supply-Side Changes in a Two-CountryEconomy with Capital Accumulations”, Journal of International Eco-nomics, 33, 127–146.

Roe, T. L., and S. Saracoglu (2004), “A Three Sector Growth Model in WhichOne Sector Is Primary”, Economic Development Center, University of Min-nesota.

Uzawa, H. (1969), “Time Preference and the Penrose Effect in a Two-ClassModel of Economic Growth”, Journal of Political Economy, 77, 628–652.

Whalley, J. (1985), Trade Liberalization among Major World Trading Areas,MIT Press, Cambridge.

Yeldan, A. E., and T. L. Roe (1994), “Environmental Policy under AlternativeMarket Structures: Lessons from Turkey as a NIC”, in Fifth InternationalCGE Modeling Conference, ed. by T. Nguyen, University of Waterloo.

34